On what to store backups: Tape VS Disk Storage
Many companies use tape archives for long-term backups and backup the most important information. It is easy to understand them: a fairly cheap, simple and reliable method of data storage, successfully used for many years - the cartridge has a shelf life of 2-3 decades, there is a lot of information on it, a streaming backup is written faster than on classical disk systems, in other words: why change something if that suits you?
It is expensive and inefficient to keep backups and backups on disk systems, and restoring any of the backups is not so often necessary, so the general slowness of the system does not bother anyone.
Fortunately, the world is not in place, technologies are developing, and today VTL (virtual tape library) has already caught up in the cost of ownership of tape archives, repeatedly surpassing them in a number of other parameters. Let's see what the tape is going to cover, and isn't it time to switch to disk libraries?
')
The tape archive is certainly a reliable and simple way to protect information, but it is not without flaws directly arising from its tape nature, mainly these difficulties are associated with the recovery of small files:
* approx.: Solved by competent software that can multi-record.
Disk array is devoid of all these disadvantages:
If I were asked to briefly describe all the advantages of disk backups, then I would not hesitate to answer: speed, reliability, scalability and flexibility.
With speed, everything is clear: reading and writing individual files from a tape is much slower than from regular hard drives. Disk systems have been evolving for a long time, they are used not only in servers, but also in ordinary desktops, and they have already accumulated rich experience in speeding up daily operations. We also considered reliability in the previous paragraph: RAID-6, physical immobility of hard drives, no need to transfer or store them in the form in which tape cartridges are stored (the cartridge can also be physically stolen during transportation, for example). But to the scalability and flexibility, I am sure there are questions, and now I will try to answer them.
I propose to consider the issue of scalability on the example of the HP StoreOnce B6200 :
The base system contains two controllers and two disk shelves with a total capacity of 48TB. Each controller can control four shelves, each with a capacity of packed railways up to 2TB. Up to eight of these controllers can be connected (3 pairs in addition to two existing ones). Thus, the B6200 will provide up to 768TB of raw capacity (due to the RAID system, the usable capacity is less by a third, but the 512TB is still an impressive figure), while its storage capacity grows as the storage capacity grows.
In this case, you are free to choose which scheme to expand the functionality of the system: first increase the volume to the limit, and then increase performance, or evenly purchase controllers with disk shelves to increase performance, and, if necessary, increase the storage capacity by installing additional disk shelves .
For the broadest backup capabilities, specialized software is responsible - HP Catalyst . HP Catalyst is a software agent that is installed on a media server (backup server) running HP DataProtector or Symantec NetBackup and Backup Exec backup software. HP Catalyst performs data deduplication right on the media servers, using the functionality of this software and sends the already deduplicated data to the HP StoreOnce system. This makes it possible to achieve high backup speeds, since several media servers are able to process a much larger stream than a single dedicated target device. For example, the HP B6200 top-end system can write data with deduplication at speeds up to 40 TB / hour, and using HP Catalyst - already up to 100 TB / hour.
The main difference between HP Catalyst and most analogs is the work not only over LAN, but also over WAN. Thus, in small regional offices, you can not install a dedicated HP StoreOnce library, but only install HP Catalyst + backup software on the media server. Next, the backup in deduplicated form will go to the HP StoreOnce library in the central office or a large territorial office. This allows multi-branch organizations to organize centralized backup management and its consolidation with minimal costs.
If you use only hardware, then for geographically distributed organizations, the consolidation of the backup is as follows. The affiliates put entry-level libraries - HP 2620, and in the center - the older model, such as HP 4430 or B6200. The branch backup is recorded on the HP StoreOnce Backup System and the data already deduplicated (20 times less than the original) are transferred to the center, where they are written to a large library. Deduplication of replicated data significantly reduces the cost of communication channels. One HP B6200 allows you to collect data from 384 branches and the entire network is managed by one administrator, which makes it possible to refuse backup administrators in branch offices. This scheme is very popular in the world, and the largest such installation in Russia already has about 100 HP StoreOnce devices and continues to grow.
We do not urge to completely abandon the time-tested technology: you can install HP Store Once as an intermediate link between user systems and a long-term archive, which will reduce the waiting time for a daily backup, carry out a partial backup of the changed parts of large files without overwriting the tape array completely, speed up the work backup and restore data, and on tape to write everything that can be useful in the long term and does not require frequent access.
It is expensive and inefficient to keep backups and backups on disk systems, and restoring any of the backups is not so often necessary, so the general slowness of the system does not bother anyone.
Fortunately, the world is not in place, technologies are developing, and today VTL (virtual tape library) has already caught up in the cost of ownership of tape archives, repeatedly surpassing them in a number of other parameters. Let's see what the tape is going to cover, and isn't it time to switch to disk libraries?
')
Tape VS Discs
The tape archive is certainly a reliable and simple way to protect information, but it is not without flaws directly arising from its tape nature, mainly these difficulties are associated with the recovery of small files:
- Significant data retrieval time;
- One application can 100% load one drive, creating backup problems for other applications; *
- The impossibility of simultaneous reading and writing if all the drives are occupied with something (it is required to wait until the operation is completed);
- The complexity of quality control and correctness of the record.
* approx.: Solved by competent software that can multi-record.
Disk array is devoid of all these disadvantages:
- Data search on the hard drive is hundreds of times faster than on the tape, which you need to find in the archive, bring, insert into the drive, rewind, start reading;
- VTL can emulate tens and hundreds of drives at once: parallel copying and data recovery for many applications without increasing the cost of system ownership;
- High reliability of data storage: server hard drives work in the most severe conditions for years, the load of the VTL system for them is not very wearing. In addition, all data is copied inside the VTL itself and protected using a RAID array, which increases both the reliability of data storage and the complexity of unauthorized access to it: even if several hard drives can be stolen, there will not be any real information on them.
Benefits of the HP StoreOnce D2D Backup System
If I were asked to briefly describe all the advantages of disk backups, then I would not hesitate to answer: speed, reliability, scalability and flexibility.
With speed, everything is clear: reading and writing individual files from a tape is much slower than from regular hard drives. Disk systems have been evolving for a long time, they are used not only in servers, but also in ordinary desktops, and they have already accumulated rich experience in speeding up daily operations. We also considered reliability in the previous paragraph: RAID-6, physical immobility of hard drives, no need to transfer or store them in the form in which tape cartridges are stored (the cartridge can also be physically stolen during transportation, for example). But to the scalability and flexibility, I am sure there are questions, and now I will try to answer them.
Scalability
I propose to consider the issue of scalability on the example of the HP StoreOnce B6200 :
The base system contains two controllers and two disk shelves with a total capacity of 48TB. Each controller can control four shelves, each with a capacity of packed railways up to 2TB. Up to eight of these controllers can be connected (3 pairs in addition to two existing ones). Thus, the B6200 will provide up to 768TB of raw capacity (due to the RAID system, the usable capacity is less by a third, but the 512TB is still an impressive figure), while its storage capacity grows as the storage capacity grows.
In this case, you are free to choose which scheme to expand the functionality of the system: first increase the volume to the limit, and then increase performance, or evenly purchase controllers with disk shelves to increase performance, and, if necessary, increase the storage capacity by installing additional disk shelves .
Flexibility
For the broadest backup capabilities, specialized software is responsible - HP Catalyst . HP Catalyst is a software agent that is installed on a media server (backup server) running HP DataProtector or Symantec NetBackup and Backup Exec backup software. HP Catalyst performs data deduplication right on the media servers, using the functionality of this software and sends the already deduplicated data to the HP StoreOnce system. This makes it possible to achieve high backup speeds, since several media servers are able to process a much larger stream than a single dedicated target device. For example, the HP B6200 top-end system can write data with deduplication at speeds up to 40 TB / hour, and using HP Catalyst - already up to 100 TB / hour.
The main difference between HP Catalyst and most analogs is the work not only over LAN, but also over WAN. Thus, in small regional offices, you can not install a dedicated HP StoreOnce library, but only install HP Catalyst + backup software on the media server. Next, the backup in deduplicated form will go to the HP StoreOnce library in the central office or a large territorial office. This allows multi-branch organizations to organize centralized backup management and its consolidation with minimal costs.
If you use only hardware, then for geographically distributed organizations, the consolidation of the backup is as follows. The affiliates put entry-level libraries - HP 2620, and in the center - the older model, such as HP 4430 or B6200. The branch backup is recorded on the HP StoreOnce Backup System and the data already deduplicated (20 times less than the original) are transferred to the center, where they are written to a large library. Deduplication of replicated data significantly reduces the cost of communication channels. One HP B6200 allows you to collect data from 384 branches and the entire network is managed by one administrator, which makes it possible to refuse backup administrators in branch offices. This scheme is very popular in the world, and the largest such installation in Russia already has about 100 HP StoreOnce devices and continues to grow.
We already have a tape backup, where to put it?
We do not urge to completely abandon the time-tested technology: you can install HP Store Once as an intermediate link between user systems and a long-term archive, which will reduce the waiting time for a daily backup, carry out a partial backup of the changed parts of large files without overwriting the tape array completely, speed up the work backup and restore data, and on tape to write everything that can be useful in the long term and does not require frequent access.
Source: https://habr.com/ru/post/186844/
All Articles